Effects of an unusual poison identify a lifespan role for Topoisomerase 2 in Saccharomyces cerevisiae

A progressive loss of genome maintenance has been implicated as both a cause and consequence of aging. Here we present evidence supporting the hypothesis that an age-associated decay in genome maintenance promotes aging in Saccharomyces cerevisiae (yeast) due to an inability to sense or repair DNA damage by topoisomerase 2 (yTop2). We describe the characterization of LS1, identified in a high throughput screen for small molecules that shorten the replicative lifespan of yeast. LS1 accelerates aging without affecting proliferative growth or viability. Genetic and biochemical criteria reveal LS1 to be a weak Top2 poison. Top2 poisons induce the accumulation of covalent Top2-linked DNA double strand breaks that, if left unrepaired, lead to genome instability and death. LS1 is toxic to cells deficient in homologous recombination, suggesting that the damage it induces is normally mitigated by genome maintenance systems. The essential roles of yTop2 in proliferating cells may come with a fitness trade-off in older cells that are less able to sense or repair yTop2-mediated DNA damage. Consistent with this idea, cells live longer when yTop2 expression levels are reduced. These results identify intrinsic yTop2-mediated DNA damage as a potentially manageable cause of aging.

K6001 is a W303-derived strain that was originally constructed for use in the study of mother cell-specific mating type switching [72,73]. K6001 encodes two integrated copies of the essential gene CDC6, independently regulated by the repressible GAL1 promoter and the mother cell-specific HO promoter, respectively (GAL1:Ub:CDC6 and HO:CDC6). Previously, we exploited this strain as an alternative to the standard microdissection method of quantifying replicative lifespan (RLS), which is labor intensive and slow [74]. When K6001 cells are grown permissively in galactose-containing media (raffinose/galactose), GAL1:CDC6 is expressed both in mother and daughter cells and the culture grows exponentially. When expression of the GAL1:CDC6 gene is repressed by glucose, only the mother-cell-specific expression of HO:CDC6 remains to support growth. Since HO expression is largely restricted to mother cells, daughters generally do not divide and growth of the culture is limited by the RLS of the initial cohort of mother cells. Since the success of this strategy depends on the conditional Death of Daughter cells, we call it the "DeaD" assay. Although promising, K6001 as a DeaD strain exhibited significant limitations. Its mother cells cease division on glucose after an average of only 3-4 divisions [74]. K6001 also has a short average lifespan under permissive conditions, presumably due to its W303 background (17 generations for K6001 [74], similar to 20.8 generations for W303R [18]). Early tests of the assay showed a reduction of DeaD lifespan by deletion of SIR2 [74] but these tests ignored the fact that the HO:CDC6 fusion is haploid-specific [75]. Deletion of SIR2 prevents silencing of the mating type information at HML and HMR and thus renders cells pseudo-diploid [76,77]. After a switch to glucose, sir2Δ mutants will thus be unable to express either copy of CDC6 (see below for mechanism of HO gene expression) and will die rapidly, as was observed [74]. When the opposite mating type information was deleted from this strain, allowing HO:CDC6 expression to sustain mother cells in glucose, no difference in DeaD lifespan was observed between SIR2 and sir2Δ cells (data not shown). Combined with the other limitations, this finding indicated that the strategy exemplified by K6001 would have to be modified before it could be used as a platform for studying replicative lifespan.
We began by switching the parental background from W303 to the longer-lived S288C background [78]. For galactose-specific CDC6 expression, we chose the strategy employed in K6001: a ubiquitin:CDC6 fusion driven by the GAL1 promoter. The N-terminal ubiquitin fusion allows modification of the amino terminus of Cdc6 to reduce protein stability and tighten control over Cdc6 activity. The ubiquitin moiety is co-translationally removed by ubiquitin processing proteases, and the amino acid serving as the new amino terminus of Cdc6 determines its half-life according to the N-end rule [79].
We cloned the GAL:Ub:CDC6 fusion from K6001 to a plasmid vector. DNA sequencing revealed a tyrosine codon at the beginning of CDC6, rather than the expected arginine [73]. Multiple independent clones from two separately obtained K6001 isolates gave the same result. According to the N-end rule, tyrosine is less destabilizing than arginine [79]. Rapid turnover of GAL1-expressed Ub-Cdc6 is required for efficient death of daughters after a shift to glucose. Arginine, which is maximally destabilizing [79], is thus the desired Nterminal residue. All of the clones also carried a conservative mutation in the ubiquitin moiety: arginine 74 to lysine. We used PCR mutagenesis to correct this mutation and to change the N-terminal residue of Cdc6 to either arginine or methionine. We then integrated these alleles, and the original K6001 allele, in place of the endogenous CDC6 in the S288C-derived strain Y7092 [80]. Since the GAL:Ub:CDC6 allele is the only source of CDC6 in these strains, all cells arrest when transferred to glucose, and the efficiency of this arrest is a function of the stability of Cdc6. Cells expressing arginine at the N-terminus of Cdc6 achieved the most efficient growth arrest. (Fig. S1A). This allele was incorporated into all subsequent DeaD strains.
The HO:CDC6 allele in K6001 is an imprecise fusion of the open reading frame of CDC6 to the HO promoter: the fusion leaves in place more than 90bp of the CDC6 5' untranslated region, and all of the 3' CDC6 untranslated sequences. This construction excludes the 3' UTR of HO, which appears to play a role in mothercell-specificity of HO expression [81]. For our new DeaD strain, we therefore created a precise replacement of the HO open reading frame with CDC6, leaving the large HO promoter and the HO 3' untranslated region in place. This HO:CDC6:HO fusion was integrated into a strain already carrying GAL1:UbR:CDC6 to create DeaD strain BB573. BB573 mother cells have improved survival on glucose compared to K6001 mother cells: mean RLS increased from 3.4 to 8.3 generations (p=5.94 x 10 -13 , Fig. S1B). However, 8.3 generations is still much shorter than the 26-28 generation mean lifespan of normal SC288c-derived cells [78].
To further improve mother cell survival in glucose, we sought to increase mother cell expression of HO:CDC6. Insight into expression of this fusion can be gained from studies of normal HO. In cells that express the wild-type HO endonuclease, mother cells switch mating type at a rate of ~70% [82]. Inhibition of switching in the remaining ~30% of mothers is dependent on the transcriptional repressor Ash1, since deletion of ASH1 increases mother cell switching to 95-100% [72,82]. These results indicate that Ash1 is normally incompletely excluded from mother cells. Ideally, Ash1 would be completely partitioned into the incipient daughter cell (the bud), leaving none in the mother cell to repress HO expression.
ASH1 is expressed at the end of mitosis, when the transcription factors Swi5 and Ace2 enter the nucleus and promote ASH1 expression by binding to four putative Swi5/Ace2 binding sites (predicted by the nucleotide sequence kGCTGr, where "K" is G or T and "R" is A or G, [83] in the ASH1 promoter. ASH1 expression is dramatically decreased in swi5 ace2 double mutants [72,84]. We deleted the Swi5/Ace2 binding sites, either in toto or in pairwise combinations, in BB573 and assessed the effect on mother and daughter cell survival on glucose by pedigree analysis (Fig. S2). Deletion of the 1 st two Swi5p binding sites (ASH1-Δ12) caused too drastic a drop in Ash1: daughter cell survival dramatically increased (represented by tall bars in Fig. S2 panel B), compare with parental strain BB573, Fig S2 panel A). This effect was caused to a lesser extent by deletion of the middle two sites (ASH1-Δ23, Fig. S2 panel C) or of all four binding sites (ASH1-Δ14, Fig S2 panel D). In contrast, deletion of the 3 rd and 4 th Swi5p binding sites produced a strain with enhanced mother cell survival without increased daughter cell "escape" (Fig S2 panel E). This strain, ASH1-Δ34, was renamed BB579 and it and its derivatives were used in all subsequent work. BB579 has a mean survival of 11.3 generations on glucose (longer than its parental strain, BB573, p=1.77 x 10 -7 ) corresponding roughly to a model combining Gompertzian senescence with a stochastic death rate of 7.5% ( Fig. S3 panel A).
Yeast cells become sterile near the end of their lifespan due to de-repression of the silent HM mating type loci and the resulting pseudo diploidy [85]. De-repression of the HM loci in a DeaD strain will result in failure to express haploid-specific HO:CDC6, as discussed above, and which might cause premature death in aging cells and an artificially short lifespan. To investigate the potential benefit of preventing pseudo diploidy by deleting one of the two mating type loci, HMR, we deleted HMR in SIR2 and sir2Δ BB579 cells. We performed microdissection (Fig. S3 panels A and B) and DeaD lifespan assays ( Fig. S3 panel C) of parental BB579 (HMR), hmrΔ, sir2Δ and hmrΔ sirΔ strains. Deletion of SIR2 alone produced an extremely short apparent RLS, presumably due to repression of HO expression resulting from pseudo diploidy, and deletion of HMR and SIR2 together gave an intermediate phenotype (Fig. S3 panels A and C) in both the microdissection and DeaD lifespan assays. None of these deletions had significant effects on permissive growth in galactose, although the hmrΔ sir2Δ strain exhibited a minor growth defect (Fig. S3 panel D). By microdissection RLS assay, deletion of HMR slightly extended RLS: mean RLS increased from 11.3 to 12.1 generations (p=1.5 x 10 -2 ), and maximum increased from 32 to 45 (Fig. S3 panel B). Deletion of SIR2 in hmrΔ BB579 cells shortened RLS to a mean of 8.6 generations (p=1.45 x 10 -4 ), with a maximum of 33 generations (Fig. S3 panel A).
The above results demonstrate that SIR2 strains have longer DeaD assay lifespans than sir2Δ strains in the improved BB579 background. To test this relationship further, and to determine whether the DeaD assay is sensitive not only to lifespan shortening but also to lifespan extension, we asked whether we could detect a range of effects on RLS by altering the expression level of SIR2 with a series of promoter fusions. Replicative lifespan varies with SIR2 expression level in yeast, since deletion of SIR2 shortens lifespan and an extra copy of SIR2 extends it [18]. There is likely to be an upper limit to lifespan extension by up-regulation of SIR2, however, since SIR2 overexpression from the GAL1 promoter causes toxicity and elevated rates of chromosome loss [86]. For our test, we used PCR-based integration to replace the endogenous SIR2 promoter with the CYC, ADH, TEF, or GPD promoter [87] in an HMR-deleted BB579 derivative. DeaD assay lifespan of the resulting strains paralleled the predicted promoter strength (CYC<ADH<TEF<GPD, [87]), with strains carrying the strongest two promoters showing extension of DeaD assay lifespan ( Fig. S4 panel C).
As expected if strong overexpression of SIR2 is toxic, the GPDpr:SIR2 fusion, which we expect to be more strongly expressed than the TEFpr:SIR2 fusion [87], confers no additional advantage for lifespan ( Fig. S4  panel C). None of these deletions had significant affects on permissive growth in galactose, although the hmrΔ sir2Δ strain exhibited a minor growth defect (Fig. S3  panel B).
In a parallel set of experiments, we replaced the endogenous SIR2 promoter with the same promoter series in a wild-type S288C strain and measured replicative lifespan using the standard microdissection assay. The results (Fig.4S Panel D) are a striking parallel to the DeaD assay findings. The TEF promoter extended mean lifespan by a robust 45%, while the stronger GPD promoter showed a lesser extension of 22%. We conclude that improvements to K6001 engineered into BB579 allow recapitulation of key features of yeast aging using high throughput capable liquid growth-based assays, including lifespan shorte-ning and extension by under-and overexpression of SIR2.